I'll buy in on scalability and cost savings. Cassandra can handle immense scale with replication and redundancy across multiple, global data centers. As for costs, we reported on one DataStax customer that spent about $250,000 to build an application that they estimated would have cost $2.5 million to build on their incumbent relational database.
But the whole idea of presenting NoSQL as a replacement for relational databases struck me as a big reach. I see NoSQL as a hugely promising platform for a wide range of never-before-possible applications -- applications where schemaless technologies are a fit -- but it doesn't logically follow that the handwriting is on the wall for relational databases.
DataStax' press release was careful not to get too carried away, but Bosworth wasn't holding back during an interview with InformationWeek.
"We now count 20 of the Fortune 100 companies among our customers, and that's how you get into Oracle's world," said Bosworth. "We're dealing with enterprises that have done their mission-critical work on Oracle for years, and suddenly they're starting to pick different technologies."
[ Want more on the future of the big data trend? Read Big Data Revolution Will Be Led By Revolutionaries. ]
It's not that Bosworth didn't have compelling arguments. He talked about how the scale of applications and the way people write them has changed, with a move away from strictly internal-facing applications with a few hundred or a few thousand users. Companies are looking beyond back-office apps, he said, and are focusing on customer-facing applications that are, almost by definition, always-available online applications with Web interfaces.
"If you're in a back office counting money, Oracle applications are great for that -- that's what ACID [atomicity, consistency, isolation, durability] compliance and relational databases are written for," said Bosworth. "But if you're writing a new application to interact with your customers, devices or sensors across the country or around the globe, you're going to have a tough time making it scale geographically and from a performance perspective without breaking your budget."
The three customers DataStax cited as examples of this purported move away from Oracle Database and MySQL were NetFlix, the well-known Internet movie streamer, OpenWave, a provider of messaging software, and Ooyala, which provides analytics, cloud-based encoding and content management tools for Internet video streaming. I would hardly characterize these as typical database customers.
Are more conventional enterprises finding new, innovative and pioneering applications that were not possible on relational databases? Absolutely. We've shared notable examples including MetLife's Facebook Wall-style view of its customers, which consolidates some 70 disparate sources of data, yet was built and deployed in little more than three months.
Is Bosworth vastly understating the continuing need for ACID-compliant relational databases that guarantee that transactions are processed reliably and in timely way? Quite obviously. NoSQL technologies have been around for decades, yet millions of organizations continue to rely on relational databases like Oracle, IBM DB2 and Microsoft SQL Server. Is the entire NoSQL community even prepared so support all those customers? DataStax, for one, currently has about 100 employees and 300 customers.
"We may be small, but we're an open source company," Bosworth responded. "The reason the technology feels so much bigger and the implications of the movement are so much bigger and the adoption is so much faster is because of the accessibility of open source and the ability to ramp up your learning in that open source ecosystem faster than we ever could back in a proprietary world."
Fine, but what about open source MySQL, which can take you a long, long way before you might have to switch to an alternative that can better handle Web scale, global applications?
"Why take yourself down a path where you know that if you're successful, you're going to have problems down the road?" Bosworth asks. "If you can start with an approach that gives you room to breathe, why would you not?"
Bosworth had some snappy comebacks, but have we really reached a tipping point? I think the "we're ready to replace Oracle" yarn was great for generating publicity. It reminds me of Cloudera's recent "unaccept the status quo" claim that the "center of gravity" is shifting toward Hadoop.
Cloudera, the biggest name in Hadoop support, is on track to become a $100 million company this year. That's a rounding error on one quarter's worth of Oracle Database business, not to mention MySQL, Microsoft SQL Server, IBM DB2 and PureData (Netezza), Teradata, EMC Greenplum, Amazon RedShift and other analytical database options.
It's absolutely the case that the NoSQL technologies and Hadoop are poised for huge growth and that we've seen the promise of new modes of building applications and managing data. We're seeing exciting and inspiring innovation that will, no doubt, dramatically change the database landscape. But it's going to take five to 10 years or more to see mass migration, so let's not get ahead of ourselves.
Yesterday’s innovative data center may be today’s money pit. Is it time for a new plan? Also in the new, all-digital Data Center Decision Time issue of InformationWeek: Data center consolidation is tough, as the government's experience shows. (Free registration required.)